AI Training and Copyright

Should writers get paid if their work is used as training material for an AI? Under current copyright laws, probably not.

Disclaimer: While I work professionally in the field of intellectual property law, I am not a lawyer. These are just my opinions about how copyright laws should work. If you have questions about how they actually work, ask a real lawyer.

Think about how a human writes. The human might go to the library and read ten books on a topic. Then the human might use what they learned in those ten books to write their own book, probably citing the ten books in their bibliography. None of the authors who wrote the ten original books got compensated in this case, except for the library’s original purchase of the books.

Now, what is the purpose of an AI? It’s to simulate a human brain. If we equate an AI with a human brain, training the AI is similar to a human learning by reading books. The AI can be trained using ten books, synthesize the information, and use the content to write its own original book using the ideas and content from the ten books. Under current copyright laws, the authors of the original works used to train the AI should not be paid anything beyond the price of the original acquisition of the content. Training an AI using copyrighted material without compensating the author is likely to be totally legal.

However, as times and technologies change, laws must also change with them. So while using copyrighted material to train an AI without author compensation is likely to be legal under our old laws, if enough people see that as unfair, then the laws can and should be changed.